You and your team are tasked with helping to determine whether an incremental new feature should be added to an existing product. The question is whether the adding the new feature will increase the number of units sold of the product. The data you have available are from other products the company sells that include this same feature.

You and your team are tasked with helping to determine whether an incremental new feature should be added to an existing product. The question is whether the adding the new feature will increase the number of units sold of the product. The data you have available are from other products the company sells that include this same feature.

Because your company is a global company, it is able to release new features to different parts of the world at different times. You determine that a simple analysis that could be done would be to look across all the company’s markets and compare units sold between those markets that have the new feature and those that don’t adjusting for potentially confounding differences between the various markets.

An analyst on your team conducts an associational analysis and fits a number of different models to the data. Across the primary and secondary models, she notes that the estimate of the change in the number of units sold associated with the inclusion of the new feature ranges from an increase of 20% to an increase of 21%, which everyone agrees is a narrow range, and the uncertainty associated with those estimates is roughly the same across the models.

Which model should be chosen as the final model, to be used in future stages of the data analysis?